RE: svn commit: r36383 - in trunk/subversion: libsvn_subr tests/libsvn_subr

> -----Original Message-----
> From: Greg Stein [mailto:gstein_at_gmail.com]
> Sent: Saturday, March 07, 2009 2:15 PM
> To: dev_at_subversion.tigris.org
> Subject: Re: svn commit: r36383 - in trunk/subversion: libsvn_subr
> tests/libsvn_subr
>
> On Sat, Mar 7, 2009 at 13:34, Bert Huijben <rhuijben_at_sharpsvn.net>
> wrote:
> >...
> > @@ -932,6 +933,15 @@ test_dirent_is_canonical(const char **ms
> > tests[i].path,
> > canonical ? "TRUE" : "FALSE",
> > tests[i].canonical ? "TRUE" :
> "FALSE");
> > +
> > + canonicalized = svn_dirent_canonicalize(tests[i].path, pool);
> > +
> > + if (canonical && (strcmp(tests[i].path, canonicalized) != 0))
> > + return svn_error_createf(SVN_ERR_TEST_FAILED, NULL,
> >...
> > @@ -1035,6 +1049,16 @@ test_uri_is_canonical(const char **msg,
> > tests[i].path,
> > canonical ? "TRUE" : "FALSE",
> > tests[i].canonical ? "TRUE" :
> "FALSE");
> > +
> > + canonicalized = svn_uri_canonicalize(tests[i].path, pool);
> > +
> > + if (canonical != (strcmp(tests[i].path, canonicalized) == 0))
> > + return svn_error_createf(SVN_ERR_TEST_FAILED, NULL,
>
> Why are these two checks different?

        Thanks for noticing,

Currently the checks pass in both formats.. but I don't think we should check that canonicalizing a path can always fix a dirent when it was not canonical before..

In almost every case we can, but I don't think we can ever convert "a:::" into something sensible on Windows.

The current simple uri rules say we always can.. But I planned to apply the same change there, but I forgot to do that before committing.

        Bert

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1283732
Received on 2009-03-07 17:38:27 CET